Patterns and risk factors for locoregional failures after mastectomy for breast cancer: an International Breast Cancer Study Group report

(1)

Annals of Oncology 23: 2852–2858, 2012 doi:10.1093/annonc/mds118 Published online 9 July 2012

Patterns and risk factors for locoregional failures after

mastectomy for breast cancer: an International Breast

Cancer Study Group report

P. Karlsson

1

_{*, B. F. Cole}

2,3

_{, B. H. Chua}

4

_{, K. N. Price}

3,5

_{, J. Lindtner}

6

_{, J. P. Collins}

7

_{, A. Kovács}

8

_,

B. Thürlimann

9

_{, D. Crivellari}

10

_{, M. Castiglione-Gertsch}

11

_{, J. F. Forbes}

12

_{, R. D. Gelber}

3,5,13

_,

A. Goldhirsch

14,15

_{& G. Gruber}

16

_{for the International Breast Cancer Study Group}

1

Department of Oncology, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Sahlgrenska University Hospital, Gothenburg, Sweden;

2

Department of Mathematics and Statistics College of Engineering and Mathematical Sciences, University of Vermont, Burlington;3

IBCSG Statistical Center, Department of Biostatistics and Computational Biology, Dana-Farber Cancer Institute, Boston, USA;4

Department of Radiation Oncology, Peter MacCallum Cancer Centre and University of Melbourne, Melbourne, Australia;5

Frontier Science and Technology Research Foundation, Boston, USA;6

The Institute of Oncology, Ljubljana, Slovenia;7

Department of Surgery, Royal Melbourne Hospital, Victoria, Australia;8

Department of Pathology, Sahlgrenska University Hospital, Gothenburg, Sweden;9

The Breast Center, Kantonsspital, St Gallen, Switzerland and Swiss Group for Clinical Cancer Research (SAKK), Bern, Switzerland;10

Department of Medical Oncology, Centro di Riferimento Oncologico, Aviano, Italy;11

Gyneco-Oncology Unit, University Hospital, Geneva, Switzerland;12

Australian New Zealand Breast Cancer Trials Group, University of Newcastle, Calvary Mater Newcastle, Newcastle, Australia;13

Harvard School of Public Health and Harvard Medical School, Boston, USA;14

European Institute of Oncology, Milan, Italy;

15_{Swiss Center for Breast Health, Sant’Anna Clinics, Lugano-Sorengo;}16

Institut fuer Radiotherapie, Klinik Hirslanden, Zürich, Switzerland

Received 4 August 2011; revised 4 January 2012; accepted 20 March 2012

Background:

Rates and risk factors of local, axillary and supraclavicular recurrences can guide patient selection and target for postmastectomy radiotherapy (PMRT).

Patients and methods:

Local, axillary and supraclavicular recurrences were evaluated in 8106 patients enrolled in 13 randomized trials. Patients received chemotherapy and/or endocrine therapy and mastectomy without radiotherapy. Median follow-up was 15.2 years.

Results:

Ten-year cumulative incidence for chest wall recurrence of >15% was seen in patients aged <40 years (16.1%), with≥4 positive nodes (16.5%) or 0–7 uninvolved nodes (15.1%); for supraclavicular failures >10%: ≥4 positive nodes (10.2%); for axillary failures of >5%: aged <40 years (5.1%), unknown primary tumor size (5.2%), 0–7 uninvolved nodes (5.2%). In patients with 1–3 positive nodes, 10-year cumulative incidence for chest wall recurrence of >15% were age <40, peritumoral vessel invasion or 0–7 uninvolved nodes. Age, number of positive nodes and number of uninvolved nodes were signiﬁcant parameters for each locoregional relapse site.

Conclusion:

PMRT to the chest wall and supraclavicular fossa is supported in patients with≥4 positive nodes. With 1–3 positive nodes, chest wall PMRT may be considered in patients aged <40 years, with 0–7 uninvolved nodes or with vascular invasion. Theﬁndings do not support PMRT to the dissected axilla.

Key words: adjuvant treatment, breast cancer, locoregional recurrence, postmastectomy radiotherapy

introduction

Postmastectomy radiotherapy (PMRT) in patients with breast

cancer reduces the risk of locoregional recurrence (LRR) by a

proportionate 60%

–70% [

1 ], and improvement in locoregional

control can impact overall survival (OS). The Early Breast

Cancer Trialists’ Collaborative Group (EBCTCG) overview

showed that one breast cancer death would be avoided for

every four LRR prevented [

1 ]. The reported absolute risk of

LRR after mastectomy without radiotherapy (RT) varies widely

[

2 ] and thus conclusions regarding indications for PMRT are

not uniform. Furthermore, the 4 : 1 ratio of LRR to OS

reported by the EBCTCG may differ among patient subgroups.

Kyndi et al. [

3 ,

4 ] reported a larger translation of LRR

reduction into survival beneﬁt in patients with more favorable

prognostic factors, e.g. in hormone receptor-positive patients

compared with hormone receptor-negative or HER-2-positive

patients.

In addition to the indications for PMRT, there is also

controversy about the optimal radiation target volume. A

recent survey on the radiotherapeutic management of invasive

breast cancer in North America and Europe found marked

differences in physician opinions. For example, internal

mammary chain irradiation was offered more often by

*Correspondence to: Dr P. Karlsson, Department of Oncology, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Sahlgrenska University Hospital, Per Dubbsgatan 14, 5thﬂoor, Gothenburg S 413 45, Sweden. Tel: +46-31-342-7401; Fax: +46-31-820114; E-mail: per.karlsson@oncology.gu.se

(2)

European than North American radiation oncologists, whereas

those from North America were more likely to irradiate the

supraclavicular fossa and axilla [

5 ].

In a previous report from the International Breast Cancer

Study Group (IBCSG) 1138 LRR were recorded among 5352

breast cancer patients treated with mastectomy without PMRT

and followed for a median of 14.5 years: an overall LRR rate of

21.3%. Among these, the most common site of LRR in breast

cancer patients without PMRT was the chest wall (53%),

followed by the supra/infraclavicular region (26%) and the

axilla (13%). Tumor relapse at the internal mammary region

was rarely reported (1%) [

6 ]. The number of positive axillary

nodes and tumor grade were signiﬁcant risk factors for LRR. In

addition, peritumoral vessel invasion (PVI) for premenopausal

patients and tumor size for postmenopausal patients were also

signiﬁcant prognostic factors [

6 ]. In a subsequent IBCSG

study, the number of examined uninvolved nodes was found to

be a signi

ﬁcant risk factor for LRR [

7 ]. In this study, patients

with 1

–3 positive nodes in the presence of PVI, young age or

few uninvolved nodes had an increased risk of LRR.

The aim of the present analysis is to study the rates and risk

factors of local, axillary and supraclavicular recurrences

separately, not only to guide patient selection for PMRT but also

to guide radiation target volume when radiotherapy is indicated.

patients and methods

design of the studies

IBCSG Trials I–IX and 11–14 (13 trials in total) accrued 12 409 patients from 1978 to 1999. Results of the treatment comparisons, detailed deﬁnitions for menopausal status, patient characteristics and eligibility have been described elsewhere [8–17] and are summarized in Supplementary Table S1 (available at Annals of Oncology online). With the exception of trial V, where patients were entered before the pathological work-up was completed, patients were included only if the tumors were pT1, pT2 or pT3 and the resection margins were free of tumor cells. Study guidelines required axillary dissection and that at leastﬁve (trials I–IV) or eight (trials V–IX and 11–14) lymph nodes should be removed in the axillary specimen. All patients on trials I–V were to receive mastectomy without RT. Patients on trials VI–IX and 11–14 received either mastectomy without RT or breast-conserving surgery, mostly with RT. Patients treated with breast-conserving surgery were excluded from the present analysis. The 13 randomized trials evaluated the timing and/or duration of chemotherapy, endocrine therapy, chemoendocrine therapy or no adjuvant therapy. Institutional review boards reviewed and approved the protocols and informed consent was required according to the criteria established within the individual countries.

For trials I–V, VIII–IX and 11–14, a central pathology review process included the histological evaluation of primary tumor specimens for invasion of lymphatic or blood vessels around the primary tumor was undertaken [18]. No central pathology review was conducted for trials VI– VII, and the information about vessel invasion was provided by the local pathology work-up from the participating centers. PVI was defined as the presence of tumor cell emboli within a vessel space, which were identified by associatedfibrin clot and/or an endothelial cell lining. The study protocol required that at least two sections of primary tumor be taken at right angles to one another to include the interface of the growing tumor border and the adjacent breast tissue. Generally,∼6 cm2of breast tissue immediately adjacent to the primary tumor, but within 1 cm of the tumor border, was available for the assessment of PVI.

patient selection

As in our previous reports [6,7], the population studied comprised patients assigned to receive total mastectomy without RT but with adequate systemic treatment, deﬁned as three or more courses of classical cyclophosphamide, methotrexate, and 5-ﬂuouricil (CMF) for

premenopausal node-positive patients, three or more courses of CMF or tamoxifen for 1–5 years for postmenopausal node-positive patients, chemotherapy and/or endocrine therapy for pre- and postmenopausal node-negative patients or endocrine therapy alone for patients with hormone receptor-positive tumors. These criteria resulted in the exclusion of 768 patients who were assigned lesser adjuvant therapy and 3441 patients who received breast-conserving surgery. An additional 82 patients who had PMRT were also excluded. Twelve patients were excluded due to missing information regarding number of uninvolved nodes. The total number of patients excluded was 4303, and the remaining 8106 patients were included in this analysis.

statistical analysis

The following variables were deﬁned for the analysis: nodal involvement status (0, 1–3, ≥4 lymph nodes involved), tumor size (≤2 or >2 cm), estrogen receptor (ER) status (<10 or≥10 fmol/mg of cytosol protein, or, in later years, based on immunohistochemical results), age (<40, 40–49, 50–59, ≥60 years), histological grade (1, 2, 3) and PVI (yes or no). To account for missing values, we included an additional category (unknown). The number of uninvolved lymph nodes was deﬁned as the number of nodes examined less the number of positive nodes.

This analysis considered the following four types of recurrences: local, axillary, supraclavicular and distant. Internal mammary recurrences were rarely reported (<1%) and were not included in this analysis [6]. Only the ﬁrst documented recurrence, possibly in combination with other sites, was considered.

The time to recurrence was determined as the number of years from randomization until thefirst proven recurrence or the date of last follow-up (or death). If no recurrence or death was documented, then time to recurrence was censored at the date of last follow-up. Statistical methods for competing risks were used in this analysis, including cumulative incidence estimation [19,20] and competing risks regression analysis [21]. Results from competing risk regression analyses were converted to hazard ratios. Analysis for each recurrence type was carried out separately. For each type of recurrence, all other failures, including death, were treated as competing risks. When evaluating a particular recurrence type, only those other types of recurrence not in combination with the type of interest were considered competing risks. Cumulative incidence curves were compared using the method of Gray [20]. Wald tests [22] were used to determine statistical significance of each risk factor in the regression models, first for each risk factor overall and then for each individual hazard ratio. A two-sided P value <0.05 was considered statistically significant. To account for multiple comparisons in the regression models, we considered a hazard ratio to be statistically significant if its two-sided P value was <0.05 and the corresponding overall Wald test had a P value <0.05.

The inﬂuence of number of uninvolved nodes on recurrence risk was descriptively evaluated using a subpopulation treatment effect pattern plot (STEPP) analysis [23] in which patients were divided into overlapping subgroups based on the number of uninvolved nodes. Each subgroup was designed to contain at least 200 patients and to overlap with the previous subgroup by at most 100 patients. The 10-year cumulative incidence of recurrence was determined within each subgroup, and the results were plotted on a graph (versus the midpoint of the interval) to illustrate how risk changes as the number of uninvolved nodes increases.

(3)

For the group of patients with 1–3 positive nodes, risk proﬁles considering age, PVI and number of uninvolved nodes were modeled, giving estimates of 10-year cumulative incidence for local, axillary and supraclavicular relapse separately.

results

Table

1 summarizes the patients and tumor characteristics of

the 8106 eligible patients. The median follow-up for all patients

was 15.2 years and the trial-speci

ﬁc median follow-up ranges

from 9.7 to 25.3 years. The 10-year cumulative incidence of

local, axillary and supraclavicular recurrence in different

patients groups is also shown in Table

1 . In general, the

absolute level of LRR was highest for the chest wall: a 10-year

cumulative incidence for local failure of >15% was documented

in patients below age 40 (16.1%), in

≥4 positive lymph nodes

(16.5%) and in patients with 0–7 uninvolved lymph nodes

(15.1%). In regard to supraclavicular failure, only patients with

≥4 positive lymph nodes exceeded a 10% cumulative risk level

(10.2%). Axillary failure rates were relatively rare with a risk far

below 10%. A 10-year cumulative risk of

∼5% could be

demonstrated for patients below age 40 (5.1%),

≥4 positive

lymph nodes (4.9%), patients with unknown tumor size (5.2%)

and patients with 0–7 uninvolved nodes (5.2%). Details are

given in Table

1 .

The cumulative incidence of distant failure exceeded the

incidence of all sites of LRR in both positive and

node-negative patients (Figure

1 ).

Multivariable competing risk regression analyses for local,

axillary and supraclavicular recurrences are shown in Table

2 .

Age, number of positive lymph nodes and number of

uninvolved lymph nodes were highly signiﬁcant parameters for

all individual locoregional relapse sites. The risk for

supraclavicular recurrence was lower in patients with positive

ER status [hazard ratio = 0.71, 95% con

ﬁdence interval (CI)

0.58–0.87, Table

2 ] compared with negative ER status. Larger

tumor size and PVI were signiﬁcant predictors for local und

supraclavicular failure but only of borderline signi

ﬁcance for

axillary relapse. Tumor differentiation, especially grade 3, was a

highly signiﬁcant predictor for supraclavicular failure (Table

2 ).

A descriptive STEPP analysis shows a decreased risk of local,

axillary, supraclavicular as well as distant recurrences with

increasing number of uninvolved nodes examined

(Supplementary Figure S1, available at Annals of Oncology

online).

Looking speci

ﬁcally at patients with 1–3 positive nodes, risk

pro

ﬁles considering age, PVI and number of uninvolved nodes

were modeled giving estimates of 10-year cumulative incidence

separately for chest wall, axillary and supraclavicular

recurrences (Supplementary Table S2, available at Annals of

Oncology online). All patients <40 years at diagnosis and

almost all groups of patients with PVI or 0–7 uninvolved

nodes had >10% 10-year cumulative incidence for chest wall

relapse (Supplementary Table S2, available at Annals of

Oncology online). For axillary recurrences, only patients <40

years combined with PVI and low numbers of uninvolved

nodes had 10-year axillary relapse rates above 5%. For

supraclavicular relapse, the combination of PVI and low

numbers of uninvolved nodes had relapse rates above 5%.

discussion

In our study of 8106 patients treated with mastectomy without

RT, the absolute LRR rates at the chest wall, axilla and

supraclavicular fossa varied. Generally, the risk factors for

locoregional failure at any site were also risk factors for the

individual anatomical subsites with the exception of ER status.

Positive ER status was found to be associated with an increased

risk of recurrence at the chest wall but not the supraclavicular

fossa. This may be a chance

ﬁnding, as a study by Kyndi et al.

reported a greater survival beneﬁt of PMRT in patients with

ER-positive disease, in which a prevented isolated chest wall

recurrence might be more meaningful in diminishing the risk

for subsequent distant spread of the disease.

Table 1. Ten-year cumulative incidence of local recurrence, axillary recurrence and supraclavicular recurrence

Risk factor Local Axillary Supraclavicular

No. (%) of patients percent (SE) percent (SE) percent (SE) Age, years <40 949 (12) 16.1 (1.2) 5.1 (0.7) 6.3 (0.8) 40–49 2607 (32) 10.5 (0.6) 2.7 (0.3) 6.1 (0.5) 50–59 2452 (30) 9.6 (0.6) 2.4 (0.3) 6.3 (0.5) ≥60 2098 (26) 10.8 (0.7) 2.5 (0.3) 3.5 (0.4) Nodes involved None 2555 (32) 6.8 (0.5) 1.3 (0.2) 2.2 (0.3) 1–3 3260 (40) 10.3 (0.5) 2.6 (0.3) 4.8 (0.4) 4–10 1744 (22) 15.4 (0.9) 4.9 (0.5) 8.8 (0.7) ≥11 547 (7) 19.9 (1.7) 4.9 (0.9) 14.8 (1.5) Tumor size, cm ≤2 3200 (39) 8.7 (0.5) 2.4 (0.3) 3.5 (0.3) >2 4623 (57) 12.4 (0.5) 3.0 (0.3) 7.0 (0.4) Unknown 283 (3) 12.9 (2.1) 5.2 (1.4) 3.8 (1.2) Tumor grade 1 1126 (14) 8.2 (0.8) 1.3 (0.4) 2.0 (0.4) 2 3520 (43) 10.7 (0.5) 2.6 (0.3) 4.3 (0.3) 3 3036 (37) 12.3 (0.6) 3.5 (0.3) 8.4 (0.5) Unknown 424 (5) 11.1 (1.6) 3.8 (0.9) 4.5 (1.0)

Estrogen receptor status

Negative 2383 (29) 10.6 (0.6) 3.1 (0.4) 7.7 (0.5) Positive 4760 (59) 11.3 (0.5) 2.5 (0.2) 4.4 (0.3)

Unknown 963 (12) 10.0 (1.0) 3.6 (0.6) 5.5 (0.7)

Peritumoral vessel invasion

No 3823 (47) 8.6 (0.5) 2.0 (0.2) 3.8 (0.3) Yes 2754 (34) 14.1 (0.7) 3.8 (0.4) 7.5 (0.5) Unknown 1529 (19) 11.3 (0.8) 3.3 (0.5) 6.2 (0.6) Nodes uninvolved 0–7 1925 (24) 15.1 (0.8) 5.2 (0.5) 9.3 (0.7) 8–11 1953 (24) 11.4 (0.7) 2.9 (0.4) 5.6 (0.5) 12–16 2126 (26) 9.8 (0.7) 2.2 (0.3) 4.3 (0.4) ≥17 2102 (26) 7.9 (0.6) 1.3 (0.2) 3.2 (0.4) Nodes examined ≤10 1940 (24) 12.5 (0.8) 3.8 (0.4) 5.7 (0.5) 11–14 2076 (26) 9.4 (0.6) 3.5 (0.4) 6.0 (0.5) 15–19 2053 (25) 11.7 (0.7) 2.2 (0.3) 5.3 (0.5) ≥20 2037 (25) 10.3 (0.7) 2.0 (0.3) 5.1 (0.5)

(4)

In the Danish and Canadian randomized trials, which

demonstrated a breast cancer-speciﬁc survival advantage of

PMRT, the radiation target volume included both the chest

wall and regional lymph nodes in the axilla, supraclavicular

fossa and internal mammary chain [

24 –

26 ]. However, it is

unclear if comprehensive locoregional RT as prescribed in the

Danish and Canadian trials is essential for the survival

improvement or RT to a more limited target volume may

achieve comparable outcome. Controversial reports about the

LRR rates and the potential harmful effects of large-ﬁeld

comprehensive locoregional irradiation such as cardiovascular

morbidity, lymphedema, pneumonitis and brachial plexopathy

[

1 ,

27 ,

28 ] leave this issue unresolved.

It is well accepted that patients with

≥4 positive nodes

should receive PMRT to the chest wall [

29 ]. Several guidelines

also advise additional irradiation of the supraclavicular lymph

nodes in these patients [

30 ] (http://www.nccn.org/

professionals/physician_gls/PDF/breast.pdf; http://www.

senologie.org/download/pdf/s3_leitlinie_en.pdf ). In our study,

patients with

≥4 positive nodes had the highest 10-year

cumulative incidence of local (16.1%) and supraclavicular

failure rates (10.2%). This

ﬁnding is consistent with the study

by Strom et al. [

31 ], who found a 15% risk for supraclavicular

failure at 10 years in patients with

≥4 positive nodes after

mastectomy. Furthermore, we observed the highest hazard

ratio (3.28, 95% CI 2.37–4.53) for supraclavicular failure in

patients with

≥4 positive nodes compared with node-negative

patients, which was highly statistically signi

ﬁcant. These

ﬁndings support the inclusion of supraclavicular nodes in

PMRT for patients with

≥4 positive lymph nodes.

In contrast, failures in the dissected axilla were uncommon

with the reported failure rates of

∼3% [

6 ,

31 ]. In our study, the

10-year cumulative incidence of axillary recurrence ranged

from 1.3% (grade 1 tumor, pN0 or

≥17 uninvolved nodes) to

∼5% (≥4 positive nodes; unknown tumor size age below 40 or

0 –7 uninvolved nodes). Therefore, our data supports the

recommendations in many radiotherapy guidelines not to

irradiate the dissected axilla. This recommendation is also

appropriate for patients with positive nodes and extracapsular

tumor spread (10-year axillary failure rates 3.2% and 4.9% in

patients with 1

–3 and ≥4 positive nodes, respectively) [

32 ].

The most controversial aspect of PMRT is its impact on

patients with 1–3 positive lymph nodes. The EBCTCG

overview showed that PMRT reduced LRR rates (5-year

absolute gain of 16.1%) and resulted in a statistically signi

ﬁcant

improvement of breast cancer mortality (15-year absolute gain

of 8.1%, P = 0.001) in these patients [

33 ]. Despite this level I

evidence, the necessity for PMRT in patients with 1–3 positive

nodes remains contentious because the overview data are

mainly driven by the Danish trials, which had a number of

extensively discussed therapeutic weaknesses. Seventy percent

of the St Gallen expert panel did not routinely recommend

PMRT in these patients but 72% would support its application

in the presence of additional risk factors including young age

or PVI [

29 ]. The retrospectively constructed risk proﬁles for

patient with 1

–3 positive nodes in our study found a 10-year

risk level for chest wall recurrence of >10% in patients aged

<40 years; PVI was present; or there were fewer than eight

uninvolved nodes. This

ﬁnding may indicate that at least chest

wall radiotherapy should be considered in these groups, which

is also supported by data from MacDonald et al. [

34 ].

Furthermore, only patients with both PVI and fewer than eight

uninvolved nodes had a relapse risk level in the supraclavicular

fossa of >5% and only patients with a combination of all three

risk factors of age <40 years, PVI and fewer than eight

uninvolved nodes had a relapse rate in the axilla of >5%. These

observations may inform appropriate radiation target volume

but should be interpreted with caution due to the retrospective

character of our study. We used age, PVI and number of

uninvolved nodes in the risk pro

ﬁles since they were the main

parameters for LRR in patients with 1–3 positive nodes and

dividing the proﬁles further would result in small numbers of

patients in each subgroup. Other studies yielded similar

supraclavicular recurrence rates ranging from 1% to 5% in

patients with 1–3 positive nodes [

32 ,

35 –

39 ]. Yu et al. [

40 ]

found lymphovascular invasion, extracapsular extension and

numbers and levels of involved axillary nodes as prognostic

factors for supraclavicular relapse and concluded that patients

with two or more of these risk factors might beneﬁt from

supraclavicular RT. The actively recruiting SUPREMO trial

(http://www.supremo-trial.com) evaluates the role of PMRT in

patients with 1

–3 positive nodes. A recently reported phase III

Figure 1. Cumulative incidence of local, axillary, supraclavicular and distant recurrence according to nodal status at diagnosis for node-negative (A), 1–3 positive nodes (B) and≥4 positive node (C) subpopulations.

(5)

randomized trial by Whelan et al. [

41 ] shows a signiﬁcant

disease-free survival advantage of breast plus regional nodal

RT compared with breast RT alone in patients treated with

breast-conserving therapy of whom 85% had 1

–3 positive

nodes. However, whether this

ﬁnding is equally applicable for

selected patient subgroups that had a mastectomy remains to

be con

ﬁrmed.

Our study found a reduced risk of any type of LRR with

increasing number of examined uninvolved nodes

(Supplementary Figure S1). This

ﬁnding is in line with other

studies [

42 ,

43 ]. Although the number of uninvolved nodes

Table 2. Full model multivariable Analysis of local recurrence (A), axillary recurrence (B), and supraclavicular recurrence (C)

Risk factor Hazard ratio (95% CI) P valuea

A. Local recurrence Age, years <40 1.00 <0.0001 40–49 0.68 (0.56–0.83) 0.0001 50–59 0.63 (0.51–0.77) <0.0001 ≥60 0.70 (0.57–0.86) 0.0009 Nodes involved None 1.00 <0.0001 1–3 1.34 (1.11–1.60) 0.0017 4–10 1.85 (1.51–2.27) <0.0001 ≥11 2.10 (1.60–2.74) <0.0001 Tumor size, cm ≤2 1.00 0.012 >2 1.22 (1.06–1.40) 0.0066 Unknown 1.43 (0.99–2.07) 0.059 Tumor grade 1 1.00 0.31 2 1.14 (0.92–1.42) 0.22 3 1.23 (0.99–1.54) 0.063 Unknown 1.13 (0.78–1.64) 0.53

Negative 1.00 0.084

Positive 1.18 (1.01–1.37) 0.039

Unknown 1.02 (0.81–1.29) 0.84

No 1.00 <0.0001 Yes 1.40 (1.21–1.62) <0.0001 Unknown 1.08 (0.90–1.31) 0.42 Nodes uninvolved 0–7 1.00 0.014 8–11 0.96 (0.80–1.15) 0.63 12–16 0.90 (0.75–1.09) 0.30 ≥17 0.73 (0.60–0.89) 0.0023 B. Axillary recurrence Age, years <40 1.00 0.0054 40–49 0.59 (0.41–0.85) 0.0043 50–59 0.53 (0.36–0.77) 0.0010 ≥60 0.57 (0.38–0.85) 0.0061 Nodes involved None 1.00 0.0024 1–3 1.66 (1.09–2.52) 0.019 4–10 2.44 (1.52–3.92) 0.0002 ≥11 1.90 (1.06–3.41) 0.031 Tumor size, cm ≤2 1.00 0.087 >2 0.98 (0.74–1.30) 0.88 Unknown 1.82 (1.03–3.22) 0.039 Tumor grade 1 1.00 0.053 2 1.58 (0.96–2.58) 0.070 3 1.94 (1.18–3.19) 0.0090 Unknown 1.92 (0.97–3.78) 0.060

Negative 1.00 0.83

Continued

Table 2. Continued

Risk factor Hazard ratio (95% CI) P valuea

Positive 0.98 (0.73–1.31) 0.87

Unknown 1.10 (0.73–1.64) 0.65

No 1.00 0.055 Yes 1.41 (1.04–1.91) 0.027 Unknown 1.06 (0.73–1.55) 0.75 Nodes uninvolved 0–7 1.00 <0.0001 8–11 0.69 (0.49–0.97) 0.031 12–16 0.55 (0.38–0.80) 0.0018 ≥17 0.34 (0.21–0.53) <0.0001 C. Supraclavicular recurrence Age, years <40 1.00 0.0007 40–49 1.12 (0.83–1.51) 0.47 50–59 1.24 (0.92–1.68) 0.16 ≥60 0.71 (0.50–1.00) 0.049 Nodes involved None 1.00 <0.0001 1–3 1.94 (1.42–2.64) <0.0001 4–10 3.02 (2.17–4.20) <0.0001 ≥11 4.39 (2.95–6.55) <0.0001 Tumor size, cm ≤2 1.00 0.015 >2 1.35 (1.09–1.67) 0.0063 Unknown 0.87 (0.44–1.72) 0.68 Tumor grade 1 1.00 <0.0001 2 1.57 (1.05–2.35) 0.030 3 2.57 (1.72–3.84) <0.0001 Unknown 1.63 (0.84–3.16) 0.15

Negative 1.00 0.0035

Positive 0.71 (0.58–0.87) 0.0008

Unknown 0.83 (0.60–1.13) 0.24

No 1.00 0.034 Yes 1.34 (1.07–1.67) 0.0095 Unknown 1.17 (0.88–1.54) 0.27 Nodes uninvolved 0–7 1.00 0.0014 8–11 0.86 (0.67–1.11) 0.25 12–16 0.77 (0.59–1.02) 0.064 ≥17 0.54 (0.39–0.74) 0.0001 a

Overall Wald test P values are shown in italics for each risk factor.

(6)

may reﬂect individual anatomic variability, fewer uninvolved

nodes examined might be associated with inadequate surgery

or pathological understaging, which could lead to local or

systemic undertreatment. Other studies have shown that the

risk for LRR decreases with the total number of lymph nodes

examined [

2 ] and the nodal ratio ( proportion of lymph nodes

examined that contain tumor) is correlated to the risk for LRR

[

44 ]. We have made a separate multivariable analysis in which

the number of uninvolved nodes is replaced by the total

number of examined nodes with very similar results (data not

shown). In the analyses for a former publication from our

group [

7 ], we investigated a number of approaches to the

statistical modeling including the use of nodal ratios and found

that grouping patients into quartiles of positive nodes and

uninvolved nodes resulted in improved model

ﬁt.

To conclude, in this analysis of 13 IBCSG randomized trials

involving over 8000 patients, the chest wall is the most

common site of locoregional failure sites after mastectomy.

Patients who were aged <40 years, had

≥4 positive nodes or

had 0–7 uninvolved nodes experienced a 10-year cumulative

incidence for local failure of

≥15%. PMRT to the chest wall

should be considered in these patients. Our study also supports

the application of supraclavicular RT in patients with

≥4

positive nodes. Irradiation of the dissected axilla is not

indicated. No clear recommendation on PMRT could be given

for patients with 1

–3 positive nodes but PMRT to the chest

wall may be considered in the presence of the risk factors of

young age (<40 years), PVI or few uninvolved nodes (0–7).

acknowledgements

We thank the patients, physicians, nurses and data managers

who participate in the International Breast Cancer Study

Group trials.

funding

This work was supported (in part) by Swiss Group for Clinical

Cancer Research; Frontier Science and Technology Research

Foundation; The Cancer Council Australia; Australian New

Zealand Breast Cancer Trials Group (National Health Medical

Research Council); National Institutes of Health (CA-75362);

Swedish Cancer Society; Cancer Association of South Africa;

Foundation for Clinical Cancer Research of Eastern

Switzerland (OSKK).

disclosure

The authors have declared no con

ﬂicts of interest.

references

1. Clarke M, Collins R, Darby S et al. Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005; 366: 2087–2106. 2. Taghian A, Jeong JH, Mamounas E et al. Patterns of locoregional failure in

patients with operable breast cancer treated by mastectomy and adjuvant chemotherapy with or without tamoxifen and without radiotherapy: results from

ﬁve National Surgical Adjuvant Breast and Bowel Project randomized clinical trials. J Clin Oncol 2004; 22: 4247–4254.

3. Kyndi M, Sorensen FB, Knudsen H et al. Estrogen receptor, progesterone receptor, HER-2, and response to postmastectomy radiotherapy in high-risk breast cancer: the Danish Breast Cancer Cooperative Group. J Clin Oncol 2008; 26: 1419–1426.

4. Kyndi M, Overgaard M, Nielsen HM et al. High local recurrence risk is not associated with large survival reduction after postmastectomy radiotherapy in high-risk breast cancer: a subgroup analysis of DBCG 82 b&c. Radiother Oncol 2009; 90: 74–79.

5. Ceilley E, Jagsi R, Goldberg S et al. Radiotherapy for invasive breast cancer in North America and Europe: results of a survey. Int J Radiat Oncol Biol Phys 2005; 61: 365–373.

6. Wallgren A, Bonetti M, Gelber RD et al. Risk factors for locoregional recurrence among breast cancer patients: results from International Breast Cancer Study Group Trials I through VII. J Clin Oncol 2003; 21: 1205–1213.

7. Karlsson P, Cole BF, Price KN et al. The role of the number of uninvolved lymph nodes in predicting locoregional recurrence in breast cancer. J Clin Oncol 2007; 25: 2019_–2026.

8. Castiglione-Gertsch M, Johnsen C, Goldhirsch A et al. The International (Ludwig) Breast Cancer Study Group Trials I-IV: 15 years follow-up. Ann Oncol 1994; 5: 717–724.

9. Ludwig Breast Cancer Study Group. Combination adjuvant chemotherapy for node-positive breast cancer. Inadequacy of a single perioperative cycle. N Engl J Med 1988; 319: 677–683.

10. Ludwig Breast Cancer Study Group. Prolonged disease-free survival after one course of perioperative adjuvant chemotherapy for node-negative breast cancer. N Engl J Med 1989; 320: 491–496.

11. International Breast Cancer Study Group. Duration and reintroduction of adjuvant chemotherapy for node-positive premenopausal breast cancer patients. J Clin Oncol 1996; 14: 1885–1894.

12. International Breast Cancer Study Group. Effectiveness of adjuvant chemotherapy in combination with tamoxifen for node-positive postmenopausal breast cancer patients. J Clin Oncol 1997; 15: 1385–1394.

13. International Breast Cancer Study Group. Endocrine responsiveness and tailoring adjuvant therapy for postmenopausal lymph node-negative breast cancer: a randomized trial. J Natl Cancer Inst 2002; 94: 1054_–1065.

14. International Breast Cancer Study Group. Adjuvant chemotherapy followed by goserelin versus either modality alone for premenopausal lymph node-negative breast cancer: a randomized trial. J Natl Cancer Inst 2003; 95: 1833–1846. 15. Thurlimann B, Price KN, Gelber RD et al. Is chemotherapy necessary for

premenopausal women with lower-risk node-positive, endocrine responsive breast cancer? 10-year update of International Breast Cancer Study Group Trial 11-93. Breast Cancer Res Treat 2009; 113: 137_–144.

16. International Breast Cancer Study Group. Toremifene and tamoxifen are equally effective for early-stage breast cancer:ﬁrst results of International Breast Cancer Study Group Trials 12-93 and 14-93. Ann Oncol 2004; 15: 1749–1759. 17. International Breast Cancer Study Group. Effects of a treatment gap during

adjuvant chemotherapy in node-positive breast cancer: results of International Breast Cancer Study Group (IBCSG) Trials 13-93 and 14-93. Ann Oncol 2007; 18: 1177_–1184.

18. Davis BW, Gelber RD, Goldhirsch A et al. Prognostic signiﬁcance of peritumoral vessel invasion in clinical trials of adjuvant therapy for breast cancer with axillary lymph node metastasis. Hum Pathol 1985; 16: 1212–1218.

19. Kalb_{ﬂeish JD, Prentice RL. The Statistical Analysis of Failure Time Data.} New York, NY: Wiley 1980.

20. Gray RJ. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988; 16: 1141–1154.

21. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999; 94: 496_–509.

22. Fleming TR, Harrington DP. Counting Processes and Survival Analysis. New York: Wiley 1991.

23. Lazar AA, Cole BF, Bonetti M, Gelber RD. Evaluation of treatment-effect heterogeneity using biomarkers measured on a continuous scale: Subpopulation Treatment Effect Pattern Plot. J Clin Oncol 2010; 28: 4539–4544.

(7)

24. Overgaard M, Hansen PS, Overgaard J et al. Postoperative radiotherapy in high-risk premenopausal women with breast cancer who receive adjuvant

chemotherapy. Danish Breast Cancer Cooperative Group 82b Trial. N Engl J Med 1997; 337: 949–955.

25. Overgaard M, Jensen MB, Overgaard J et al. Postoperative radiotherapy in high-risk postmenopausal breast-cancer patients given adjuvant tamoxifen: Danish Breast Cancer Cooperative Group DBCG 82c randomised trial. Lancet 1999; 353: 1641–1648.

26. Ragaz J, Jackson SM, Le N et al. Adjuvant radiotherapy and chemotherapy in node-positive premenopausal women with breast cancer. N Engl J Med 1997; 337: 956–962.

27. Pierce S, Recht A, Lingos T et al. Long-term radiation complications following conservative surgery (CS) and radiation therapy (RT) in patients with early stage breast cancer. Int J Radiat Oncol Biol Phys 1992; 23: 915–923.

28. Lingos T, Recht A, Vicini F et al. Radiation pneumonitis in breast cancer patients treated with conservative surgery and radiation therapy. Int J Radiat Oncol Biol Phys 1991; 21: 355–360.

29. Goldhirsch A, Ingle JN, Gelber RD et al. Thresholds for therapies: highlights of the St Gallen International Expert Consensus on the primary therapy of early breast cancer 2009. Ann Oncol 2009; 20: 1319_–1329.

30. Aebi S, Davidson T, Gruber G et al. Primary breast cancer: ESMO Clinical Recommendations for Diagnosis, Treatment and Follow-up. Ann Oncol 2010; 21Suppl 5v9–v14.

31. Strom E, Woodward WA, Katz A et al. Clinical investigation: regional nodal failure patterns in breast cancer patients treated with mastectomy without radiotherapy. Int J Radiat Oncol Biol Phys 2005; 63: 1508–1513.

32. Gruber G, Cole BF, Castiglione-Gertsch M et al. Extracapsular tumor spread and the risk of local, axillary and supraclavicular recurrence in node-positive, premenopausal patients with breast cancer. Ann Oncol 2008; 19: 1393–1401.

33. Darby S. Overview of randomised trials of radiotherapy in early breast cancer. Cancer Res 2009; 69 (Suppl 24) (Abstr MS3–1).

34. Macdonald SM, Abi-Raad RF, Alm El-Din MA et al. Chest wall radiotherapy: middle ground for treatment of patients with one to three positive lymph

nodes after mastectomy. Int J Radiat Oncol Biol Phys 2009; 75: 1297–1303.

35. Vicini FA, Horwitz EM, Lacerna MD et al. The role of regional nodal irradiation in the management of patients with early-stage breast cancer treated with breast-conserving therapy. Int J Radiat Oncol Biol Phys 1997; 39: 1069–1076. 36. Recht A, Pierce SM, Abner A et al. Regional nodal failure after conservative

surgery and radiotherapy for early-stage breast carcinoma. J Clin Oncol 1991; 9: 988–996.

37. Recht A, Gray R, Davidson NE et al. Locoregional failure 10 years after mastectomy and adjuvant chemotherapy with or without tamoxifen without irradiation: Experience of the Eastern Cooperative Oncology Group. J Clin Oncol 1999; 17: 1689–1700.

38. Livi L, Scotti V, Saieva C et al. Outcome after conservative surgery and breast irradiation in 5,717 patients with breast cancer: implications for supraclavicular nodal irradiation. Int J Radiat Oncol Biol Phys 2010; 76: 978–983.

39. Truong PT, Jones SO, Kader HA et al. Patients with T1 to T2 breast cancer with one to three positive nodes have higher local and regional recurrence risks compared with node-negative patients after breast-conserving surgery and whole-breast radiotherapy. Int J Radiat Oncol Biol Phys 2009; 73: 357–364. 40. Yu JI, Park W, Huh SJ et al. Determining which patients require irradiation of the

supraclavicular nodal area after surgery for N1 breast cancer. Int J Radiat Oncol Biol Phys 2010; 78: 1135–1141.

41. Whelan TJ, Olivotto I, Ackerman I et al. NCIC-CTG MA.20: an intergroup trial of regional nodal irradiation in early breast cancer. J Clin Oncol 2011; 29 (18 suppl) LBA1003.

42. Vinh-Hung V, Cserni G, Burzykowski T et al. Effect of the number of uninvolved nodes on survival in early breast cancer. Oncol Rep 2003; 10: 363–368. 43. Salama JK, Heimann R, Lin F et al. Does the number of lymph nodes examined

in patients with lymph node-negative breast carcinoma have prognostic signi_{ﬁcance?. Cancer 2005; 103: 664–671.}

44. Truong PT, Woodward WA, Thames HD et al. The ratio of positive to excised nodes identiﬁes high-risk subsets and reduces inter-institutional differences in locoregional recurrence risk estimates in breast cancer patients with 1-3 positive nodes: an analysis of prospective data from British Columbia and the M. D. Anderson Cancer Center. Int J Radiat Oncol Biol Phys 2007; 68: 59–65.